Structural basis of substrate recognition by a bacterialdeubiquitinase important for dynamics ofphagosome ubiquitinationMichael J. Sheedloa,1, Jiazhang Qiub,1, Yunhao Tanb,1,2, Lake N. Paulc, Zhao-Qing Luob,3, and Chittaranjan Dasa,3

aDepartment of Chemistry, Purdue University, West Lafayette, IN 47907; bDepartment of Biological Sciences, Purdue University, West Lafayette, IN 47907;and cBindley Biosciences Center, Purdue University, West Lafayette, IN 47907

Edited by Brenda A. Schulman, St. Jude Children’s Research Hospital, Memphis, TN, and approved November 3, 2015 (received for review July 27, 2015)

Manipulation of the host’s ubiquitin network is emerging as animportant strategy for counteracting and repurposing the post-translational modification machineries of the host by pathogens. Ubiq-uitin E3 ligases encoded by infectious agents are well known, as are avariety of viral deubiquitinases (DUBs). Bacterial DUBs have been dis-covered, but little is known about the structure and mechanism un-derlying their ubiquitin recognition. In this report, we found thatmembers of the Legionella pneumophila SidE effector family harbora DUB module important for ubiquitin dynamics on the bacterialphagosome. Structural analysis of this domain alone and in com-plex with ubiquitin vinyl methyl ester (Ub-VME) reveals unique mo-lecular contacts used in ubiquitin recognition. Instead of relying onthe Ile44 patch of ubiquitin, as commonly used in eukaryotic coun-terparts, the SdeADub module engages Gln40 of ubiquitin. The ar-chitecture of the active-site cleft presents an open arrangementwith conformational plasticity, permitting deubiquitination of threeof the most abundant polyubiquitin chains, with a distinct preferencefor Lys63 linkages. We have shown that this preference enables effi-cient removal of Lys63 linkages from the phagosomal surface. Re-markably, the structure reveals by far the most parsimonious use ofmolecular contacts to achieve deubiquitination, with less than 1,000Å2 of accessible surface area buried upon complex formation withubiquitin. This type of molecular recognition appears to enable dualspecificity toward ubiquitin and the ubiquitin-like modifier NEDD8.

type IV secretion | ubiquitination | Legionella | phagosome |deubiquitinase

Ubiquitin, a small, 76-aa protein modifier, is involved in awide array of eukaryotic cellular processes. The function-

ality of ubiquitin depends on the precise timing of the conjugation/deconjugation of the C terminus of ubiquitin to the e-amino groupof a lysine residue of a target protein. At the heart of this processare ligases (responsible for the covalent attachment of ubiquitin)and deubiquitinases (DUBs), which function to cleave isopeptidebonds between ubiquitin and substrates or within polyubiquitinchains (1). Even though many eukaryotic DUBs have already beencharacterized, little is known of these enzymes in prokaryotes (1–3).Given the essential role of ubiquitination in eukaryotic cells, it

is not surprising that infectious agents have evolved numerous el-egant strategies to exploit host signaling mediated by ubiquitination.Many bacterial pathogens use virulence factors to hijack the hostubiquitin pathway to establish successful infections (4). Even thoughE3 ubiquitin ligases of bacterial or viral origin have been relativelywell characterized, bacterial DUBs have not, despite their impor-tance in the life cycles and pathogenicity of several microbial species,including Salmonella typhimurium (SseL), Chlamydia trachomatis(ChlaDub1 and ChlaDub2), and ElaD (Escherichia coli) (5–7). Thesebacterial DUBs belong to a larger group of peptidases called the CEclan (the MEROPS database), which comprises eukaryotic, bacterial,and viral representatives (7, 8). Although quite different fromeukaryotic DUBs (<10% identity), bacterial DUBs are phyloge-netically related to mammalian desumoylating enzymes (SENP1,2, and 3) and the deneddylase Den1 (NEDP1 or SENP8). In light

of a large divergence from their eukaryotic counterparts, the overallstructure of these proteases, as well as how they function and rec-ognize ubiquitin to act as DUBs, remains to be structurallyanalyzed.Modulation of host ubiquitination pathways has emerged as an

important theme in the pathogenesis of the opportunistic pathogenLegionella pneumophila responsible for Legionnaires’ disease (4, 9).Ubiquitinated species are enriched on the L. pneumophila-con-taining vacuole (LCV), and interference with such associa-tion disturbs bacterial intracellular replication (10). Among theapproximately 300 effectors injected into host cells by L. pneu-mophila via the Dot/Icm type IV secretion system (11), eight proteinsappear to possess F-box or U-box domains typical of some E3 ligases(12–16). This ligase activity has been demonstrated for LegU1,LegAU13/AnkB, and LubX (14, 17). A recent study revealed thatSidC and SdcA are E3 ligases that catalyze the ligation reactionwith a unique mechanism, and are required for efficient enrich-ment of ubiquitinated species on the bacterial phagosome (18).Because balanced regulation of host cell processes is critical for

the virulence of L. pneumophila (19), we initiated experiments toidentify L. pneumophila proteins with DUB activity. Our effortsrevealed that members of the SidE family contain a DUB domain,which catalyzes the reaction with a Cys-His-Asp (CHD) catalytictriad showing a preference for Lys63-linked polyubiquitin chains.Structural analysis of the DUB domain and its complex with themechanism-based inhibitor, ubiquitin vinyl methyl ester (Ub-VME),

Significance

Ubiquitination and deubiquitination have emerged in recentyears as novel targets for the design of therapeutic agents. Toour knowledge, the structure of the deubiquitinase (DUB) do-main of SdeA represents the first prokaryotic DUB determinedand will thus potentially serve as a model for other bacterialdeubiquitinating enzymes for use in structure-guided drugdesign. Legionella pneumophila ubiquitin E3 ligases play im-portant roles in the biogenesis of the phagosome permissivefor bacterial replication. The discovery of effectors with DUBactivity highlights the importance of modulation of host pro-cesses in a regulated and balanced manner.

Author contributions: M.J.S., J.Q., Y.T., Z.-Q.L., and C.D. designed research; M.J.S., J.Q., Y.T.,and C.D. performed research; L.N.P. contributed new reagents/analytic tools; M.J.S., J.Q.,Y.T., L.N.P., Z.-Q.L., and C.D. analyzed data; and M.J.S., Z.-Q.L., and C.D. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Data deposition: The atomic coordinates and structure factors have been deposited in theProtein Data Bank, www.pdb.org (PDB ID codes 5CRA, 5CRB, and 5CRC).1M.J.S., J.Q., and Y.T. contributed equally to this work.2Present address: Division of Gastroenterology, Boston Children’s Hospital and HarvardMedical School, Boston, MA 02115.

3To whom correspondence may be addressed. Email: luoz@purdue.edu or cdas@purdue.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1514568112/-/DCSupplemental.

15090–15095 | PNAS | December 8, 2015 | vol. 112 | no. 49 www.pnas.org/cgi/doi/10.1073/pnas.1514568112

Dow

nloa

ded

by g

uest

on

Apr

il 9,

202

1

revealed a canonical core ubiquitin-like protease (Ulp) fold witha ubiquitin interface that is quite different from those used bystructurally characterized eukaryotic DUBs. We also found thatalthough the DUB activity is dispensable for the SidE family’srole in intracellular bacterial replication, it is important for thedynamics of the association of ubiquitinated species with thebacterial phagosome.

ResultsIdentification of SdeA as a Deubiquitinating Enzyme. To identifyL. pneumophila proteins with potential DUB activity, we used thesuicide inhibitor HA-Ub-VME, capable of forming a covalentcomplex with proteins harboring active DUB domains with a cat-alytic cysteine (20). When the probe was used to tag DUBs inlysates of L. pneumophila, a single band with a molecular weightlarger than 150 kDa was detected (Fig. 1A). Because the CHDmotif is most commonly used in DUB catalysis, we analyzedL. pneumophila effectors near the molecular weight of the iden-tified band for this motif, and found that it is present in membersof the SidE family (Fig. S1). This reactivity was then confirmedin vitro by labeling all three putative DUB domains with HA-Ub-VME as evidenced by a ∼8-kDa shift corresponding to theformation of a ubiquitin adduct (Fig. 1B). We then showed thatfull-length SdeA and the DUB domains of SdeA, SdeB, andSdeC could disassemble the most common chain types (Lys11,Lys48, Lys63; Fig. S2 A–D) (21).To measure the activity of SdeA1–193, referred to as SdeADub,

against polyubiquitin chains attached to a protein, we prepared apolyubiquitinated GFP substrate (GFP-Ubn) with mostly Lys63-linked chains and small populations of chains linked via Lys11 andLys48 (22, 23). SdeADub was able to completely deubiquitinate thissubstrate within 10 min (Fig. 1C). We further determined the ac-tivity of SdeADub against more biologically relevant substrates. Tothis end, we created a cell line stably expressing HA-Ubiquitin.Ubiquitinated proteins were then immunoprecipitated with an HA-specific antibody. His6-SdeADub almost completely removed theubiquitin signals, whereas the Cys118-Ala catalytically inactivemutant could not (Fig. 1D).To explore whether SdeADub recognizes ubiquitin-like (Ubl)

posttranslational modifications, we probed for its reactivity withother Ubl proteins (NEDD8, ISG15, and SUMO) conjugated tovinyl sulfone (VS). In these assays, SdeADub was found to alsoform a covalent complex with NEDD8-VS and, to a lesser extent,ISG15-VS. There was no detectable reactivity with the SUMO-VS probe (Fig. 1E). The ability to form the complex requiredCys118, further supporting the notion that members of the SidEfamily harbor an active DUB domain. Consistent with reactivitywith NEDD8-VS, SdeADub was able to remove NEDD8 signalsfrom neddylated proteins prepared from mammalian cellstransfected to express Flag-NEDD8 (Fig. 1F).

SdeADub Is Involved in Polyubiquitin Regulation on the BacterialPhagosome. Members of the SidE family are similar proteins withsizes larger than 150 kDa (24, 25). Mutants lacking all members ofthe family are defective in intracellular replication in the amoebaehost Dictyostelium discoideum (25). We then determined the role ofDUB activity in the function of the SidE family in infections. At72 h post infection, a mutant lacking all four members of the SidEfamily (Δ4) displayed approximately 100-fold defects in D. dis-coideum, with the defect fully complemented by expressing SdeA(Fig. S3A). Interestingly, the Cys118-Ala mutant is still fully func-tional in the complementation of the defects (Fig. S3A), suggestingthat DUB activity is not required for the role of the SidE paralogs inintracellular bacterial replication.Phagosomes harboring virulent L. pneumophila are known to

be enriched with polyubiquitinated species (10); we then de-termined the role of DUB activity in such enrichment. In mousebone marrow-derived macrophages, ∼40% of the vacuoles

containing WT L. pneumophila stained positively for ubiquitin(Fig. 2A). The rate of ubiquitin-positive vacuoles increased toapproximately 90% in infections using the Δ4 strain; such in-crease can be reduced to nearly WT levels by expressing SdeAbut not the SdeAC118A mutant (Fig. 2B). Similar reduction inubiquitin-positive vacuoles is also seen with the expression of othermembers of the SidE family (Fig. S4 A and B). The mutant wastranslocated into host cells indistinguishably to that of similarlyexpressed WT protein (Fig. S3 B and C), indicating that its inability

Fig. 1. Members of the SidE family harbor a domain with DUB activity.(A) Lysates of L. pneumophila cells incubated with HA-Ub-VME for 1 h at 37 °Cand immunoprecipitated with HA-specific beads. Protein samples were re-solved by SDS/PAGE and probed with an HA-specific antibody. (B) Each DUBprotein (Top, SdeA/B/CDub; Bottom, full-length SdeA) was incubated withHA-Ub-VME. (C) Recombinant SdeA/B/CDub was incubated with Ubn-GFP for10 min at room temperature. The complete removal of polyubiquitin fromthe substrate is observed based on the appearance of GFP. (D) Cleavage ofubiquitin chains by SdeADub on polyubiquitinated proteins. Ubiquitinatedproteins were isolated from a cell line stably expressing HA-Ub, incubatedwith SdeADub or SdeADub Cys118Ala, and probed with an HA-specific antibody.(E) Reactivity of SdeADub with Ubl modifiers. His6-SdeADub was incubated withUb/Ubl-VS probes. (F) The cleavage of NEDD8 chains by SdeADub on neddylatedproteins. Recombinant SdeADub or SdeADub Cys118Ala was incubated with pro-teins immunoprecipitated from cells transfected to stably express Flag-NEDD8.Neddylation signals were probed with the Flag antibody.

Sheedlo et al. PNAS | December 8, 2015 | vol. 112 | no. 49 | 15091

BIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

Apr

il 9,

202

1

to complement was caused by loss of enzymatic activity in theDUB module. Together, these results reveal that the DUB ac-tivity of SdeA is important for the dynamics of ubiquitin associ-ation with the bacterial phagosome.

SdeADub Cleaves Lys11/48/63 Chains with a Preference for Lys63. Todetermine the chain type preference of SdeADub, we used a gel-based diubiquitin cleavage assay. These results revealed that, eventhough SdeADub cleaves all chain types examined (Fig. 3A and Fig.S2E), there was a clear preference for Lys63-linked chains, yieldinga binding constant (from the Km) of ∼200 μM for each DUB do-main and ∼100 μM for the full-length SdeA construct (Fig. 3B andFig. S2H). We were unable to obtain kinetic parameters for Lys11and Lys48-linked diubiquitin chains as a result of the low cleavageefficiency. Interestingly, high concentrations of Lys48-linked diubi-quitin appeared to inhibit SdeADub activity (Fig. S2 F and G). Inline with the biochemical observations, the SdeA DUB module wasable to efficiently remove Lys63-linked polyubiquitin chains fromthe phagosomal surface (Fig. S4 C and D).

Structure of the SdeA DUB Module. Structures of SdeADub weredetermined for two constructs: 1–163 SeMet (apo), 1–193 Cys118-Ala (apo), and 1–193-Ub-VME at 2.85 Å, 2.02 Å, and 2.53 Å, re-spectively, with each solution containing two molecules in theasymmetric unit (even though biophysical analysis suggests thisprotein is a monomer in solution; see Fig. S7 F and G and TableS1). The overall fold of SdeADub can be described as an arrange-ment of periodic secondary structures forming two distinct lobes,the α-lobe and the β-lobe (Fig. 4A and Fig. S5A), which pack closelyto generate a compact globular fold, connected by loops L2 and L10(Fig. 4A and Fig. S5A). The β-lobe is composed of a twisted, mixed,six-stranded β-sheet (β1–β6; Fig. S5A), which, when viewed fromone edge, appears to form a partially wound β-barrel (Fig. S5B).The concave surface of this partial β-barrel allows a curved α-helix(α3) to make tight contact. An additional, relatively short, helix (α2)seems to cap one end of the partial barrel. The outer surface of thebarrel packs against two α-helices (α1 and α4) of the α-lobe (Fig.S5A). The rest of the helices in the α-lobe (α5, α6, and α7) arrangethemselves by packing against at least one other helix withinthis lobe.The active site is located at the interface between the two

lobes, with Cys118 placed in the first turn of α4 facing the cat-alytic histidine, His64, with the catalytic aspartic acid, Asp80,within short hydrogen-bonding distance (Fig. 4B). This CHDcatalytic triad is complemented by a suitably located Asn residue

(Asn114) in loop L2 that we propose stabilizes the oxy-anion species(Fig. 4B). The catalytic residues line an open catalytic cleft (exceptfor Glu9, which seems to function as a flap) with a predominantlynegative electrostatic character complementing the distinctly elec-tropositive C-terminal tail of ubiquitin and some Ubls. Interestingly,inspection of the structures of the three constructs reported hereinreveals that the active-site residues have the propensity to assumedifferent conformations, indicated by different orientations of theCys and His residues, possibly reflecting plasticity in the catalyticresidues, which, in addition to the open nature of the cleft, mayconfer some advantage in accommodating substrates with varyingstructure of their polyubiquitin chain.

SdeADub Binds Ubiquitin Through Three Distinct Patches. Sequencealignments and modeling with other Ulps revealed little aboutthe nature of ubiquitin binding. To visualize the mode of ubiquitinrecognition, we prepared a covalent complex of SdeADub withubiquitin by reacting the DUB module with Ub-VME. The struc-ture of the complex reveals the formation of a covalent bondbetween Cys118 and the VME group of the suicide inhibitor.Inspection of the Ub-VME bound crystal structure reveals threedistinct ubiquitin binding sites: the active-site cleft (site-1), a sitecentered around Tyr33 (site-2) referred to as the Y-site, and a

Fig. 2. The SdeA DUB module is important for regulating the dynamics of ubiquitin association with the bacterial phagosome. (A) Mouse bone marrow-derived macrophages were infected with the indicated L. pneumophila strains (LP03 is the Dot/Icm-deficient strain; Δ4 refers to a strain lacking all SidE familymembers) for 1 h, and fixed cells were differentially stained for extra- and intracellular bacteria. Ubiquitinated species were labeled with the FK2 antibody.Images were acquired with a fluorescence microscope. (Scale bar: 5 μm.) (B) Quantification of the enrichment of ubiquitinated species associated with the LCVs.Samples in triplicate were inspected for the enrichment of ubiquitin signals with the bacterial phagosomes. At least 150 vacuoles were scored for each sample.

Fig. 3. Comparison of cleavage efficiency of diubiquitin substrates. (A) Relativeefficiency of Lys11, Lys48, and Lys63 diubiquitin cleavage by SdeADub. Each assaywas done in triplicate, with error bars representing SDs. (B) Kinetic analysis ofSdeA (blue) and SdeADub (green). Fitting the plots to the Michaelis–Mentenequation yields Km values of ∼75 μM and ∼200 μM for the full length and DUBdomain, respectively.

15092 | www.pnas.org/cgi/doi/10.1073/pnas.1514568112 Sheedlo et al.

Dow

nloa

ded

by g

uest

on

Apr

il 9,

202

1

site that engages the β1–β2 loop of ubiquitin (site-3; Figs. 4B and 5).The active-site cleft engages the tetrapeptide portion of the C-terminal tail of ubiquitin (Leu73-Gly-VME, with Gly-VME corre-sponding to Gly76) with contacts that involve mainly interbackbonehydrogen bonding. Hydrogen-bonding contacts with tail residuesstabilize the tail in an extended conformation, resulting in an ar-rangement in which the tail residues form the central strand of athree-stranded β-sheet that includes one arm of loop L6 and resi-dues Glu9–Gln12 of the enzyme (Fig. 4B). Engagement of the tail isfacilitated by charge and shape complementarity, with the α-carbonof Gly-VME fitting snugly into a narrow groove at the end of theactive-site cleft adjacent to Cys118. This matching of shape ensuresproper placement of the scissile isopeptide bond adjacent to Cys118and ensures selectivity for the terminal di-Gly motif in ubiquitin andcertain Ubls. Arg72, a key determinant of ubiquitin/Ubl specificity,is unengaged, explaining the reactivity of the DUB toward ubiq-uitin-VS and NEDD8-VS (Fig. 1E) (26, 27).Most DUB-ubiquitin cocrystal structures show active-site tail

binding as observed here, albeit with more extensive interactions.Tyr33 in the Y-site makes three contacts with ubiquitin, the mostprominent of which appears to be Gln40, as well as making van derWaals contacts with Ile36 and Leu73. The hydroxyl group is en-gaged in a tight hydrogen-bonding interaction with the side-chaincarbonyl group of Gln40, whose side-chain amide proton serves as ahydrogen bond donor to Asp61 from the catalytic cleft (Fig. 5C).Another commonly observed interaction between DUBs andubiquitin involves the β1–β2 loop of ubiquitin engaged at a distinctsite that allows the dipeptide turn to fit in snugly. In SdeADub, the

β1–β2 loop of ubiquitin interacts primarily through van der Waalscontacts, with only one hydrogen bond between the backbone car-bonyl of Leu8 of ubiquitin and the hydroxyl group of Ser29 of SdeA(Fig. 5D). It appears that most of these interactions can occur evenif the turn segment has substitutions such as those found in ISG15,possibly explaining the observed reactivity with ISG15-VS andNEDD8-VS (Fig. 1E and Fig. S6). Surprisingly, the ubiquitin Ile44patch is not engaged in binding (Fig. 5A), quite distinct fromeukaryotic DUBs, which seem to rely on the Ile44 patch for affinity,and also from the Ulp family member Den1’s recognition ofNEDD8 (1, 27, 28).

Structural Similarity and Differences from Den1. Structural similarityanalysis using the DALI server (29) identified Den1 (NEDP1 orSENP8) apo and in complex with NEDD8 as high-scoring hits[Protein Data Bank (PDB) ID codes 2BKR, 2BKQ, and 1XT9](27, 28). Den1 is a member of the Ulp family of Cys proteases,whose members share a characteristic core fold comprised of afive-stranded mixed β-sheet packed against a helix (27, 28, 30).Den1 and SdeADub are structurally similar, with most of theβ-sheets (excluding β6) and α4 (bearing the catalytic cysteine)adopting a similar fold (Fig. S7A). Even though the core Ulpdomains were consistent, differences exist, particularly in regionsused for binding to ubiquitin or Ubls (Fig. S7B). A simple S-shapedloop at the beginning of SdeADub is replaced in Den1 by a helixfollowed by a β-hairpin and is used in NEDD8 binding. Notably,β2 of SdeA, which forms a hairpin turn with β1 to create the Tyr33binding loop, is replaced with α3 in Den1, which mediates NEDD8interaction via the Ile44 hydrophobic patch.The structures of the Den1-NEDD8 and SdeADub-Ubiquitin

complexes also exhibit differences and similarities. The mostnotable similarity is the relative orientation of the C-terminal tailwith respect to the adjacent arm of loop L6, which appears to bea characteristic element in the Ulp family. Loop L6 in our structuresis rigid, forming a β-hairpin with a bulge (Fig. S7D). In Den1, the

Fig. 4. The structure of the SdeA DUB module and in complex with Ub-VME. (A) Cartoon representation of SdeADub identifying α-helices (orange),β-sheets (yellow), and loops (black). The loops L2 and L10, which connect theα- and β-lobes, are shown in blue. (B) Ubiquitin (green) binding to SdeADub

(surface rendering) is illustrated with the catalytic residues shown as bluesticks. The C-terminal tail of ubiquitin binds to the active site (site 1). (Insets)Superimposition of the catalytic residues of SdeADub observed in threestructures reported here (Top Right), flexibility of Glu9 observed in twodistinct conformations in the apo and Ub-VME bound structures (Top Left),and arrangement of the C-terminal tail within the active site (Bottom).

Fig. 5. Molecular recognition of ubiquitin by SdeA DUB module. (A) TheC-terminal tail of ubiquitin (green) is secured in the catalytic cleft, pre-dominantly stabilized by backbone interactions (dotted lines; SdeADub resi-dues in orange). Glu6 and Glu9 participating in hydrogen bonding andelectrostatic interactions with Arg74 of ubiquitin. (B) Gln12 hydrogenbonding with the backbone nitrogen of Leu71 of ubiquitin. (C) Tyr33 andAsp61 form hydrogen bonds with Gln40 of ubiquitin. (D) Backbone atoms ofthe β1–β2 loop of ubiquitin forms hydrogen bonds with the backbone andside chain of Ser29 of SdeADub.

Sheedlo et al. PNAS | December 8, 2015 | vol. 112 | no. 49 | 15093

BIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

Apr

il 9,

202

1

corresponding loop is quite dynamic, undergoing a significantconformational change upon binding to NEDD8. SdeADub buries∼900 Å2 [according to the proteins, interfaces, structures andassemblies (PISA) server] of surface area upon binding ubiquitin,which is substantially less than Den1 (burying nearly 2,000 Å2

when bound to NEDD8) as a result of greater recognition ofNEDD8 by Den1 by structural elements absent in SdeADub (31).The most prominent of these substitutions is an N-terminalβ-hairpin in Den1 that nestles against the loop segment of NEDD8(Lys48-Asn41; Fig. S7E). This hairpin is absent in SdeADub andwould mediate the corresponding interaction with Lys48 of ubiq-uitin, making it exo-specific for this chain type (that is, capable ofcleaving ubiquitin chains only from the most distal end). Loss of thiselement in the bacterial DUB may allow it to rapidly cleave fromanywhere in the chain. The other prominent difference is the ab-sence of helix-4 (Den1 numbering) in SdeADub, which plays animportant role in engaging the Ile44 patch. To probe the impor-tance of the observed contacts in solution, we generated alaninepoint mutants and tested their activity against Lys-63–linked diu-biquitin as well as a Nedd8 fusion construct (His-MBP-NEDD8-Ub). From these experiments, we find that some variations withinthe catalytic cleft are tolerated, whereas mutating Y33 and D61(Y-site) resulted in a complete loss of activity. Moreover, resultsobserved in experiments with the NEDD8 fusion construct mir-rored those of Lys63-linked diubiquitin, suggesting a similar modeof recognition (Fig. S8).The selectivity of SdeADub for Lys63-linked chains may rely on

chain topology. Because Lys63-linked chains adopt an extendedconformation (whereas Lys48- and Lys11-linked chains do not),they may be strung more easily through the catalytic site. Basedon this observation, we hypothesize that the preference observedfor Lys63-linked chains is a result of conformational differencesbetween the chains themselves rather than structural restrictionsimposed by SdeADub. However, it is possible that selectivebinding to the amino acids that flank Lys63 contributes to thepreference for this chain type; however, this remains to be de-termined in future studies.

DiscussionPhagosomes harboring virulent L. pneumophila (i.e., LCVs) aredecorated with ubiquitinated proteins whose function is in-completely understood but is thought to be important for thebiogenesis and maintenance of the vacuole (9, 18). Earlierstudies have identified Dot/Icm effectors with ubiquitin E3 ligaseactivity with important roles in the biogenesis of LCVs (18, 32). Ourdiscovery of DUB activity associated with members of the SidEfamily of effectors points to the importance of spatiotemporalregulation of ubiquitination on the LCV surface. It adds anotherlayer of complexity to the regulation of host processes by effectorsof opposite biochemical activities exemplified by the reversibleAMPylation and phosphorylcholination of the Rab1 small GTPase(33, 34). The activity of the SidE family of effectors appears to beregulated by SidJ (35, 36), a gene within the gene cluster (37). Allmembers of this family contain the N-terminal DUB module, whichappears to redundantly regulate the extent of ubiquitination of thevacuolar surface, but their natural substrates remain to be de-termined (24). Clearly, DUB activity is not critical for the role ofSidEs in intracellular bacterial replication under laboratory in-fection conditions; such roles likely are conferred by the activity ofother domains of these large proteins (36).Our structural analysis establishes a core Ulp fold and a ca-

nonical CHD catalytic triad in a bacterial CE clan of cysteineprotease with DUB activity. The structure of the SdeADub–Ub-VME complex reveals the molecular contacts used by the DUBfor ubiquitin recognition, providing, to our knowledge, the firststructure of a Ulp-fold protease bound to ubiquitin. Bacterial CEpeptidases are predicted to share structural similarity with eukaryoticproteases such as Ulp1 (yeast) and Den1 (mammals), which exhibit

exclusive specificity toward SUMO1 and NEDD8, respectively (27,28). SdeADub is unable to recognize SUMO because of mismatchesbetween the C-terminal tail of the Ubl and the DUB, but is capableof deneddylating activity. Den1 binds to NEDD8 by using an ex-tensive interface. SdeADub, in contrast, presents a remarkably simpleinterface, burying only half the accessible surface area as a result offewer contacts with the C-terminal tail and no contacts with theLys48 loop or Ile44 patch. The latter is a feature widely used bystructurally characterized eukaryotic DUBs (27, 28). Instead,SdeADub binds to ubiquitin in an orientation that is twisted rel-ative to the way NEDD8 docks on Den1, an orientation thatbrings Gln40 of ubiquitin to the forefront and permits a bifurcatedengagement with Tyr33 and Asp61 of SdeADub (Fig. S7C). Thisdifference in binding interface between Den1 and SdeADub mayreflect their substrate preferences; Den1 releases NEDD8 from itsprecursor whereas SdeADub most likely acts on polyubiquitinchains. Although a smaller interface may result in a higher Km, asobserved here, this would not be a significant impediment becausethe enzyme is likely to encounter a relatively high local concen-tration of polyubiquitin chains on the surface of the LCV (10).The origin of ubiquitin is believed to predate that of NEDD8 and

other Ubl modifiers (7, 38). This has given rise to the speculationthat the evolutionary antecedents of the CE clan of peptidases withubiquitin/Ubl hydrolyzing activity might have originated in eukary-otes, with primarily DUB activity having been acquired by patho-genic bacteria through horizontal gene transfer (7). The knowneukaryotic members of the CE clan might have evolved to recognizeSUMO (Ulp1) or NEDD8 (Den1) exclusively at the expense ofDUB activity. The dual specificity toward ubiquitin and NEDD8 bySdeADub is quite remarkable considering that the more widely un-derstood eukaryotic CE peptidases have stringent specificity towardonly one type of Ubl modifier. Moreover, NEDD8 and ubiquitinhave respective E1 enzymes that discriminate the two based on thesingle amino acid at position 72 (26, 39). Interestingly, NEDD8binding to SdeADub appears to use a similar interface as ubiquitin,as the point mutants of Y33 and D61 abolished deneddylase ac-tivity. A binding interface with NEDD8 of the nature seen in theubiquitin-bound structure reported here can be a solution to aproblem of molecular recognition that demands both DUB anddeneddylase activity. We have shown that the SdeA DUB removesLys63-linked chains from LCVs, which could be used to antagonizethe host’s attempt to recruit autophagy machinery, reminiscent ofthe proposed function of SseL (5, 40).The suppression of SdeA toxicity by SidJ (Fig. S3D) likely reflects

an intricate interplay between the effectors of L. pneumophila toachieve spatiotemporal control of host cell processes, underscoredby a previous observation that SidJ can remove SdeA from thephagosome after a certain time postinfection (35–37). The contri-bution of the catalytic activity of the N-terminal DUB module toSidJ-mediated suppression of SdeA’s toxicity (caused by its centraldomain) adds another layer of complexity to the interplay betweenthese two effectors (Fig. S3D). It also suggests that the DUB activityof SidE DUBs may be used for multiple functions, conferring aspecific advantage to the organism in effectively establishing in-fection inside the host cell.

MethodsHEK293 cells were transfected with pCDNA3 HA-Ubiquitin or pCMV 4XFlag-NEDD8 by Lipofectamine 2000 (Life Technologies). Cells were collected 24 hafter transfection and lysed in lysis buffer [0.2% Nonidet P-40, 50 mM Tris, pH7.5, 150 mM NaCl, 1 mM EDTA, 15% (vol/vol) glycerol, and Protease InhibitorMixture Set III (EMD Millipore)]. The total cell lysates were then subjected toimmunoprecipitation by HA agarose or Flag affinity gel (Sigma-Aldrich) for 4h, and the beads were washed with lysis buffer three times and resuspendedin cleavage buffer containing 50 mM Tris, pH 7.5, 100 mM NaCl, and 1 mMDTT. Samples were divided into aliquots and left untreated or treated with100 μg of SdeADub or SdeADubCys118Ala for 1 h at room temperature. Thebeads were then washed three times with the cleavage buffer and boiled for5 min at 100 °C in 1× SDS/PAGE buffer. The proteins were then separated by

15094 | www.pnas.org/cgi/doi/10.1073/pnas.1514568112 Sheedlo et al.

Dow

nloa

ded

by g

uest

on

Apr

il 9,

202

1

SDS/PAGE, and polyubiquitinated proteins or polyneddylated proteins weredetected by HA and Flag antibodies, respectively. Additional descriptions ofmaterials and methods used in the study are provided in SI Methods.

ACKNOWLEDGMENTS. We thank Andreas Martin, Yasuke Sato, and RonaldHay for supplying constructs used in this study. We are grateful for the use ofthe Advanced Photon Source at Argonne National Laboratory at the General

Medical Sciences and Cancer Institutes Structural Biology Facility beamline23-ID-B. The Advanced Photon Source is an Office of Science User Facilityoperated for the US Department of Energy (DOE) Office of Science by ArgonneNational Laboratory and is supported by the DOE under Contract DE-AC02-06CH11357. This work was supported by National Institutes of Health Grants2R01GM103401 (to C.D.), R56AI103168 (to Z.-Q.L.), K02AI085403 (to Z.-Q.L.),and R21AI105714 (to Z.-Q.L.).

1. Komander D, Rape M (2012) The ubiquitin code. Annu Rev Biochem 81:203–229.2. Nijman SM, et al. (2005) A genomic and functional inventory of deubiquiti-

nating enzymes. Cell 123(5):773–786.3. Reyes-Turcu FE, Ventii KH, Wilkinson KD (2009) Regulation and cellular roles of

ubiquitin-specific deubiquitinating enzymes. Annu Rev Biochem 78:363–397.4. Zhou Y, Zhu Y (2015) Diversity of bacterial manipulation of the host ubiquitin

pathways. Cell Microbiol 17(1):26–34.5. Rytkönen A, et al. (2007) SseL, a Salmonella deubiquitinase required for macrophage

killing and virulence. Proc Natl Acad Sci USA 104(9):3502–3507.6. Claessen JH, et al. (2013) Catch-and-release probes applied to semi-intact cells reveal

ubiquitin-specific protease expression in Chlamydia trachomatis infection. Chembiochem14(3):343–352.

7. Catic A, Misaghi S, Korbel GA, Ploegh HL (2007) ElaD, a Deubiquitinating proteaseexpressed by E. coli. PLoS One 2(4):e381.

8. Rawlings ND, Waller M, Barrett AJ, Bateman A (2014) MEROPS: The database ofproteolytic enzymes, their substrates and inhibitors. Nucleic Acids Res 42(Databaseissue):D503–D509.

9. Hubber A, Kubori T, Nagai H (2013) Modulation of the ubiquitination machinery byLegionella. Curr Top Microbiol Immunol 376:227–247.

10. Dorer MS, Kirton D, Bader JS, Isberg RR (2006) RNA interference analysis of Legionellain Drosophila cells: Exploitation of early secretory apparatus dynamics. PLoS Pathog 2(4):e34.

11. ZhuW, et al. (2011) Comprehensive identification of protein substrates of the Dot/Icmtype IV transporter of Legionella pneumophila. PLoS One 6(3):e17638.

12. Angot A, Vergunst A, Genin S, Peeters N (2007) Exploitation of eukaryotic ubiquitinsignaling pathways by effectors translocated by bacterial type III and type IV secretionsystems. PLoS Pathog 3(1):e3.

13. Habyarimana F, et al. (2008) Role for the Ankyrin eukaryotic-like genes of Legionellapneumophila in parasitism of protozoan hosts and humanmacrophages. EnvironMicrobiol10(6):1460–1474.

14. Kubori T, Hyakutake A, Nagai H (2008) Legionella translocates an E3 ubiquitin ligasethat has multiple U-boxes with distinct functions. Mol Microbiol 67(6):1307–1319.

15. Brüggemann H, et al. (2006) Virulence strategies for infecting phagocytes deducedfrom the in vivo transcriptional program of Legionella pneumophila. Cell Microbiol8(8):1228–1240.

16. Zeng LR, Park CH, Venu RC, Gough J, Wang GL (2008) Classification, expression pat-tern, and E3 ligase activity assay of rice U-box-containing proteins. Mol Plant 1(5):800–815.

17. Ensminger AW, Isberg RR (2010) E3 ubiquitin ligase activity and targeting of BAT3 bymultiple Legionella pneumophila translocated substrates. Infect Immun 78(9):3905–3919.

18. Hsu F, et al. (2014) The Legionella effector SidC defines a unique family of ubiquitin ligasesimportant for bacterial phagosomal remodeling. Proc Natl Acad Sci USA 111(29):10538–10543.

19. Xu L, Luo ZQ (2013) Cell biology of infection by Legionella pneumophila. Microbes Infect15(2):157–167.

20. Borodovsky A, et al. (2001) A novel active site-directed probe specific for deubiqui-tylating enzymes reveals proteasome association of USP14. EMBO J 20(18):5187–5196.

21. Xu P, et al. (2009) Quantitative proteomics reveals the function of unconventionalubiquitin chains in proteasomal degradation. Cell 137(1):133–145.

22. Kim HC, Huibregtse JM (2009) Polyubiquitination by HECT E3s and the determinantsof chain type specificity. Mol Cell Biol 29(12):3307–3318.

23. Beckwith R, Estrin E, Worden EJ, Martin A (2013) Reconstitution of the 26S protea-some reveals functional asymmetries in its AAA+ unfoldase. Nat Struct Mol Biol20(10):1164–1172.

24. Luo ZQ, Isberg RR (2004) Multiple substrates of the Legionella pneumophila Dot/Icmsystem identified by interbacterial protein transfer. Proc Natl Acad Sci USA 101(3):841–846.

25. Bardill JP, Miller JL, Vogel JP (2005) IcmS-dependent translocation of SdeA intomacrophages by the Legionella pneumophila type IV secretion system. Mol Microbiol56(1):90–103.

26. Schulman BA, Harper JW (2009) Ubiquitin-like protein activation by E1 enzymes: Theapex for downstream signalling pathways. Nat Rev Mol Cell Biol 10(5):319–331.

27. Shen LN, et al. (2005) Structural basis of NEDD8 ubiquitin discrimination by thedeNEDDylating enzyme NEDP1. EMBO J 24(7):1341–1351.

28. Reverter D, et al. (2005) Structure of a complex between Nedd8 and the Ulp/Senpprotease family member Den1. J Mol Biol 345(1):141–151.

29. Holm L, Rosenstrom P (2010) Dali server: Conservation mapping in 3D. Nucleic Acids Res38(Web server issue):W545–W549.

30. Lima CD, Reverter D (2008) Structure of the human SENP7 catalytic domain and poly-SUMOdeconjugation activities for SENP6 and SENP7. J Biol Chem 283(46):32045–32055.

31. Krissinel E, Henrick K (2007) Inference of macromolecular assemblies from crystallinestate. J Mol Biol 372(3):774–797.

32. Price CT, et al. (2009) Molecular mimicry by an F-box effector of Legionella pneu-mophila hijacks a conserved polyubiquitination machinery within macrophages andprotozoa. PLoS Pathog 5(12):e1000704.

33. Neunuebel MR, et al. (2011) De-AMPylation of the small GTPase Rab1 by the path-ogen Legionella pneumophila. Science 333(6041):453–456.

34. Tan Y, Arnold RJ, Luo ZQ (2011) Legionella pneumophila regulates the small GTPaseRab1 activity by reversible phosphorylcholination. Proc Natl Acad Sci USA 108(52):21212–21217.

35. Jeong KC, Sexton JA, Vogel JP (2015) Spatiotemporal regulation of a Legionellapneumophila T4SS substrate by the metaeffector SidJ. PLoS Pathog 11(3):e1004695.

36. Havey JC, Roy CR (2015) Toxicity and SidJ-mediated suppression of toxicity require distinctregions in the SidE family of Legionella effectors. Infect Immun 83(9):3506–3514.

37. Liu Y, Luo ZQ (2007) The Legionella pneumophila effector SidJ is required for efficientrecruitment of endoplasmic reticulum proteins to the bacterial phagosome. InfectImmun 75(2):592–603.

38. Xirodimas DP (2008) Novel substrates and functions for the ubiquitin-like moleculeNEDD8. Biochem Soc Trans 36(Pt 5):802–806.

39. Souphron J, et al. (2008) Structural dissection of a gating mechanism preventingmisactivation of ubiquitin by NEDD8’s E1. Biochemistry 47(34):8961–8969.

40. Mesquita FS, et al. (2012) The Salmonella deubiquitinase SseL inhibits selectiveautophagy of cytosolic aggregates. PLoS Pathog 8(6):e1002743.

41. Swanson MS, Isberg RR (1995) Association of Legionella pneumophila with themacrophage endoplasmic reticulum. Infect Immun 63(9):3609–3620.

42. Dong KC, et al. (2011) Preparation of distinct ubiquitin chain reagents of high purityand yield. Structure 19(8):1053–1063.

43. Schneider CA, Rasband WS, Eliceiri KW (2012) NIH Image to ImageJ: 25 years of imageanalysis. Nat Methods 9(7):671–675.

44. Otwinowski Z, Minor W (1997) Processing of X-ray diffraction data collected in os-cillation mode. Methods Enzymol 276(part A):307–326.

45. Adams PD, et al. (2010) PHENIX: A comprehensive Python-based system for macro-molecular structure solution. Acta Crystallogr D Biol Crystallogr 66(pt 2):213–221.

46. Vagin A, Teplyakov A (2010) Molecular replacement with MOLREP. Acta Crystallogr DBiol Crystallogr 66(pt 1):22–25.

47. Winn MD, et al. (2011) Overview of the CCP4 suite and current developments. ActaCrystallogr D Biol Crystallogr 67(pt 4):235–242.

48. Murshudov GN, Vagin AA, Dodson EJ (1997) Refinement of macromolecular struc-tures by the maximum-likelihood method. Acta Crystallogr D Biol Crystallogr 53(pt 3):240–255.

49. Murshudov GN, Vagin AA, Lebedev A, Wilson KS, Dodson EJ (1999) Efficient aniso-tropic refinement of macromolecular structures using FFT. Acta Crystallogr D BiolCrystallogr 55(pt 1):247–255.

50. Vijay-Kumar S, Bugg CE, Cook WJ (1987) Structure of ubiquitin refined at 1.8 A res-olution. J Mol Biol 194(3):531–544.

51. Schuck P (2000) Size-distribution analysis of macromolecules by sedimentation ve-locity ultracentrifugation and lamm equation modeling. Biophys J 78(3):1606–1619.

Sheedlo et al. PNAS | December 8, 2015 | vol. 112 | no. 49 | 15095

BIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

Apr

il 9,

202

1